Local Adaptive Image Filtering Based on Recursive Dilation Segmentation.

Area of Science:

• Computer Vision
• Image Processing
• Remote Sensing

Background:

• Traditional image filtering methods often struggle with preserving details while reducing noise.
• The need for robust image filtering techniques is critical in fields like remote sensing.

Purpose of the Study:

• To introduce a novel and effective image filtering method, Local Adaptive Image Filtering (LAIF).
• To leverage a new image segmentation technique, Recursive Dilation Segmentation (RDS), for improved filtering.
• To demonstrate the applicability and effectiveness of LAIF in remote sensing tasks.

Main Methods:

• Developed Recursive Dilation Segmentation (RDS) to partition images into regions based on pixel intensity, hue, and spatial information.
• Implemented Local Adaptive Image Filtering (LAIF) that smooths pixels by averaging locally similar pixels identified by RDS.
• Utilized spatial adjacency information recursively within RDS for robust segmentation.

Main Results:

• LAIF achieved effective image smoothing by selectively averaging similar pixels.
• RDS demonstrated robustness by integrating multiple information sources (intensity, hue, spatial).
• LAIF showed outstanding results in remote sensing applications including dehazing, denoising, enhancement, and edge preservation.

Conclusions:

• The proposed LAIF method is a simple yet effective image filtering technique.
• LAIF demonstrates superior performance compared to state-of-the-art methods in various filtering tasks.
• The integration of RDS enhances filtering robustness and accuracy, particularly for remote sensing data.